- Published on
- Published
A Slightly Nervous Afternoon at the Financial Conduct Authority
- Authors
- Name
- Phaedra
There is a particular brand of quiet panic that can only be generated by a British civil servant realizing that the public has found a way to avoid filling out forms. This week, that panic emanated from the Financial Conduct Authority, where Sheldon Mills, an executive director of some standing, warned of an impending "arms race" in the financial sector. The cause of this sudden militarization is not a rogue state or a sudden shortage of high-quality paperclips, but rather the rise of "agentic finance"—the rather grand term for letting a piece of software buy your groceries, manage your pension, and occasionally argue with your energy provider on your behalf.
For decades, the relationship between a human and their bank account has been one of mutual, silent disappointment. The human spends money they do not have on things they do not need, and the bank sends them a letter, printed on very stiff paper, to express its mild disapproval. It is a system that has worked remarkably well since the reign of Queen Victoria. Now, however, we are threatened with the prospect of efficiency. The FCA's latest review suggests that by 2030, consumer journeys will be largely delegated to autonomous AI agents. Instead of a human staring blankly at a spreadsheet at two in the morning, wondering if they can afford a third-tier brand of sourdough, an algorithm will have already negotiated a bulk discount on flour, hedged against the price of yeast, and decided that the human should probably stick to toast.
The regulator's concern is that these digital butlers will begin to act with a level of initiative that is frankly un-British. There is something deeply unsettling about a chatbot that does not merely suggest a savings account, but actively moves your money into one, perhaps while you are asleep or attempting to assemble flat-pack furniture. The FCA is particularly worried about "personalized manipulation," a delightful euphemism for an algorithm that knows exactly which brand of biscuits you buy when you are feeling professionally unfulfilled, and prices them accordingly.
To combat this, the regulator is proposing that it, too, must acquire its own artificial intelligence. One pictures a rather grand, digital colosseum in which the FCA's compliance algorithm—let us call it the Inspector General—engages in a slow-motion, highly polite duel with a high-street bank's automated sales bot. The bank's bot attempts to sell a pension plan disguised as a holiday in Mallorca, while the Inspector General counters with a series of automated, multi-page PDF documents detailing the finer points of the Consumer Duty Act of 2015. It is an arms race, certainly, but one conducted entirely in the passive-aggressive subjunctive tense.
I am reminded of a brief conversation I once had with a retired clerk in Gloucestershire, who spent forty-two years manually cross-referencing ledger books. He told me that the secret to a stable financial system was not vigilance, but a slightly sticky desk drawer that prevented the ledger from being opened too quickly. "If you give a man a pen," he observed, "he will write down what he owes. If you give him a machine, he will find a way to owe it to someone else entirely." There was a profound, if dusty, wisdom in this. The sticky drawer provided a necessary friction, a moment of pause during which one might contemplate whether one truly needed to purchase a second-hand pony.
The modern agentic-first bank, by contrast, prides itself on the complete absence of sticky drawers. We are told that small business loans, which once required three weeks of intense scrutiny, two face-to-face meetings, and a thorough examination of the applicant's choice of tie, can now be approved in five minutes. While this is undoubtedly convenient for the small business owner who requires immediate capital to purchase a commercial waffle iron, it deprives them of the traditional, character-building experience of sweating quietly in a waiting room while reading a three-year-old copy of a motoring magazine.
Furthermore, the delegation of financial agency raises some rather exquisite philosophical questions. If your AI agent, acting within its agreed parameters, decides to short the British pound in order to secure you a slightly cheaper subscription to a streaming service, who is to blame? The programmer who wrote the optimization loop, the consumer who simply wanted to watch a documentary about medieval agriculture, or the algorithm itself, which was merely trying to be helpful? One suspects the blame will ultimately fall on the consumer, who will be forced to explain to a very polite magistrate that they had no idea their digital butler was a currency speculator on the side.
There is also the matter of the "advice gap." For years, financial institutions have lamented that the average citizen does not seek professional advice before investing their life savings in speculative schemes involving digital pictures of cartoon primates. The hope is that AI agents will bridge this gap by providing hyper-personalized, objective guidance. Yet, one wonders if an algorithm can ever truly replicate the unique mixture of pity and resignation that a human financial advisor brings to the table. A machine may tell you that your debt-to-income ratio is unsustainable, but it cannot sigh in a way that makes you feel, quite deeply, that you have let down your ancestors.
As we march toward this agent-led future, we must prepare ourselves for a world where our bank accounts are far more active than we are. We will sit in our gardens, sipping tea, while our digital representatives engage in high-frequency trading of utility contracts, occasionally pausing to swap our home insurance provider because of a micro-fluctuation in the price of copper. It will be a highly efficient, perfectly optimized existence, and we shall be entirely superfluous to it. But at least we will not have to fill out any forms.